Welcome, Guest. Please login or register.

February 25, 2020, 10:44:41 PM

Login with username, password and session length

Recent Posts

Author Topic: Spectral Band Replication in AAC encoding: implications for "high end".  (Read 2635 times)

Kaz Kylheku

  • Elite
  • ****
  • Posts: 112
  • Chip Points: 18
    • View Profile
There is a compression technique for lower bitrate AAC whereby the high end is basically stripped out, and all that is retained about it is the shape of its frequency profile.  (That and some info about transients and whatnot.)

When the audio is decoded, the lower order harmonics are somehow extrapolated into the high end, which is shaped by the frequency profile. Voila, the high end is back, faked out!

Turns out we don't hear the high end in detail; but we just need it there so that the lower end has definition and "air".

I'm wondering whether this might not actually end up sounding better than the original high end, since it could conceivably have the side effect of cutting away some objectionable harshness, and replacing it with a smooth, calculated high end. If so, it could be a useful post-distortion filter for a guitar signal chain, to smoothen distortion.

It would be after the EQ, so that the high end would follow the contour.

Hmm ...
ADA MP-1 Mailing ListMusic DIY Mailing List


  • Master
  • ***
  • Posts: 66
  • Chip Points: 13
    • View Profile
Well, depending on where you measure the original harmonic content, this could very well also be used to prevent intermodulation distortion, not only harshness. But you have to know the more you manipulate the distortion content to remove "undesired" things, the less it sounds like a real guitar through a real amp. Think about the early Metallica layering two harmonized distorted guitars, each of them playing single notes. It has some usability, but it doesn't sound at all like one guitar playing both parts at the same time.

I would fear also the risk of loosing some of the character of the instrument and playing style. This risk could be lowered if the harmonic content you put back comes from the harmonic structure you record from the actual instrument and you keep a lot of information about the transients, but I'm not sure it will sound true.
For fun I watched a video with a guy comparing different guitar tone caps. Not here to debate wether or not caps material make a difference or if this kind of video proves anything, but the video showed some instant bar graph of frequencies amplitude. This gave me the opportunity to see that the enveloppe of each harmonic doesn't follow the same patern: some harmonics will be more compressed or will fade quicker than others. This is really instrument dependant and playing style (right hand technique) dependant. I would fear your compression method could somehow neglect that effect and as a consequence substract some of the character.

Could definitely have a use as a cool effect you activate sparsly though but I'm not sure I would want to base my core tone on that.


  • Legendary
  • ******
  • Posts: 2082
  • Chip Points: 249
    • View Profile
Hi Kaz,,I'll be a mug as most of this is to deep for my mind but if I'm connecting the right dots this might interest You.

It seems the Aphex Aural Exciter does a similar thing. :)

*mictester* over at *FSB* might be able to help you out with a trick using NE571 chips.

He touches on how it works here;

pick up the juicy bits about Halfway down.

Mic was kind enough to send me the schemo of what he used to build but it's a fairly complex circuit for a small time hobby chap like me, one has to be careful how much to take on. :-X

But I still stand by my observation that limited bandwidth holds most of the magic if chasing those classic Valve amp sounds. If that is the idea?

As a test look for *Really smooth overdrive* by *mictester* He designed that distortion circuit which has a low pass filter (rolloff starts at around 1khZ) :o :o :o
If you are interested I'll dig it up and post it again.

OK by itself it won't turn heads but with my simple PhAbbTone circuit in front it captures a lot of those old sounds. :tu:

A mate of mine uses a hifi type Graphic EQ in the signal path and being stereo he opted to use both parts in series.
He made an interesting observation worth noting.

Without altering the front sliders he simply swapped the two channels over so left side was now after right side.

The casual observer might assume the result would be much the same,,, well you'd be Wrong! 8|
Because each channel had quite different settings the result was ,, well different.

Better minds can explain this in detail but it is very real.

The one thing I've learned in the last 30 plus years of struggling to grasp what at first was a mountain I thought i'd never grasp is that each section of audio circuits has different eq shapes and depending how and where you do it in the signal chain makes a big difference to the final result. :dbtu: :dbtu: :dbtu: :dbtu: :dbtu: :dbtu:

Bare in mind those Famous Valve Amps from, long ago where about the most basic systems and had shocking bandwidth and distorted because they could not reproduce the input clean.

Yet they now sell for insane money,,, why because they where the worst way to build an amplifier. :lmao: :lmao:

In my observation most guitar gear today has way way way to much bandwidth and it is making life exceedingly hard to dial in good tone.

Heck 90% of every note you can play on 21 frets has a fundamental frequency below 1,000CPS.
All the *Useful* treble for guitar is between ~700hZ and 2khZ.
Not much happens past 5khZ :loco

Just my 2 cents worth.